Complexity reduction of CELP coders

Abstract

The authors propose a modification of the classical CELP (code-excited linear predictive) algorithm in order to reduce its computational complexity and required memory size, while preserving the quality of the reconstructed speech. Rather than performing the individual weighting of each candidate sequence, the authors suggest a global implementation of the vocal tract weighting function at the code-book level, thanks to the use of an adaptive code-book. As a result, the analysis-by-synthesis procedure does not require the processing of all the candidate sequences through the synthesis and weighting filters, and therefore the complexity requirement of the algorithm is much reduced. The method was demonstrated on a 7.2 kb/s adaptive code-book CELP operating with the same level of quality as the 13 kb/s GSM coder which has been normalized for speech coding in the European cellular radio network. The proposed technique was then applied to the design of a coder operating with a low delay requirement.